Winners and Losers in the Labour Market of the 1990s
by A. Heisz*, A. Jackson**, G. Picot*** No. 184
11F0019 No. 184 ISSN: 1205-9153 ISBN: 0-662-31494-8
Business and Labour Market Analysis Division 24-F, R.H. Coats Building, Ottawa, K1A 0T6 *Statistics Canada (613) 951-3748 **Canadian Council on Social Development ***Statistics Canada (613) 951-8214 Facsimile Number: (613) 951-5403
March 2002
This paper represents the views of the authors and does not necessarily reflect the opinions of Statistics Canada.
Aussi disponible en français
Table of Contents Introduction ..................................................................................................................................... 1 Why might relative incomes have changed in the 1990s? .............................................................. 2 Change in trade and technology................................................................................................... 2 New employment arrangements................................................................................................... 2 The rapidly increasing human capital of Canadian workers ........................................................ 2 Changes in the institutional landscape ......................................................................................... 3 Aggregate demand during the 1990s............................................................................................ 3 Changes in income and earnings inequality .................................................................................... 4 Changes in relative earnings............................................................................................................ 8 Declining real earnings among young workers, particularly men................................................ 8 No marked increase in the wage premium for higher education in Canada among all workers, a rising premium for younger workers ....................................................... 11 The shifting relative labour market position of men and women.................................................. 13 Educational attainment is rising faster for women..................................................................... 13 The male-female earnings gap is declining................................................................................ 14 Employment indicators improving for prime aged women relative to men .............................. 17 Low-income trends in the 1990s ................................................................................................... 18 The low-income rate .................................................................................................................. 19 Canada’s low-income standing internationally.......................................................................... 21 Depth of low-income ................................................................................................................. 22 Persistence of low-income ......................................................................................................... 23 Summary ....................................................................................................................................... 25 Conclusion..................................................................................................................................... 25 Bibliography.................................................................................................................................. 27
ABSTRACT This paper examines the Canadian labour market during the 1990s and contrasts it to prior decades, with a special focus on distributional outcomes. It discusses changes in relative earnings between groups, changes in relative labour market outcomes of women and older workers, changes in earnings and income inequality, and changes in low-income.
Keywords: income inequality, earnings, low-income
Introduction The 1990s labour market is thought to have been different in many ways from earlier decades. These differences can influence income levels, distributions and the relative earnings of various groups. For example, the “ new economy” , spurred by the computer technology revolution and expanding international trade, has created new opportunities for many people, but also may have left behind less skilled workers. New employment relationships have become increasingly popular, as reflected in the rising number of workers who are self-employed. The very rapid growth in the educational attainment of workers, particularly among younger women and older workers, has certainly influenced relative earnings, and likely resulted in firms significantly altering their production processes. This again influences the relative wages of the less and more highly skilled workers through shifts in labour demand. Changes in public institutions and transfer programs has no doubt affected the incomes of less well off Canadians. Finally, all these factors occurred in the context of slow macro-economic growth, at least for the first half of the decade. Within this environment, this paper asks what has happened to the relative pay of various workers and the distribution of income among families? What groups of workers have gained and lost as a result of these changing patterns of work? The primary objective of this paper is to present results on the distribution of income in Canada, which we hope will serve as a useful introduction to this subject. This paper is descriptive in nature and draws on previously reported findings from many different sources. The paper focuses on four major topics: (1) earnings and income inequality; (2) the relative earnings of young and old, and more and less educated; (3) the changing relative position of men and women; and (4) changes in low-income in Canada during the 1990s. Earnings inequality rose modestly for men during the 1990s (compared to faster increases in the 1980s), but inequality changed little for women. For workers as a whole (men and women together), there is little evidence of rising earnings inequality, at least to 1997. However, at the family level earnings inequality did rise over the 1990s, driven largely by changes in family formation patterns. Through the middle 1990s inequality in family disposable income (after taxes and transfers) remained stable, as it had for the past few decades in Canada. Generally speaking the tax/transfer system prevented rising family earnings inequality from translating to rising family disposable income inequality, although there is evidence that family income inequality started to rise in the later 1990s. However, aggregate inequality trends often mask significant changes in relative wages. Even within the context of little change in earnings inequality, there can be significant movement in the relative wages of various groups, movements that often offset one another when added together. During the 1990s there was a continued decline in real wages among young men overall, but wages for young university educated men and women rose in the last years of the decade. The relative position of women has continued to improve in terms of labour market outcomes and wages relative to men. Low-income trends are significantly influenced by the business cycle, generally rising in recessions and falling in expansions. But during the 1990s, as the weak expansion took hold through the 1993 to 1996 period, the low-income rate did not fall as expected, but continued to rise. The result was Analytical Studies Branch – Research Paper Series
-1-
Statistics Canada No. 11F0019 No. 184
that during the latter half of the 1990s in particular, both the low-income rate (the share of the population in low-income) and the low-income gap ( the depth of low-income) was higher than comparable periods during the 1980s, at least among the non-elderly population.
Why might relative incomes have changed in the 1990s? Following are five broad factors which help to explain why incomes and work arrangements of Canadians may have changed during the 1990s. Change in trade and technology Shifts in labour demand associated with trade and technology are thought to have resulted in considerable structural change in the economy. At the core of this, workers with more skills or experience are seen to be in greater demand, while other workers are left behind. All else equal, this would result in increasing relative wages for more highly skilled workers (relative to the less skilled). If the supply of high skilled workers rose along side demand, then the impact of increased demand for high skilled workers on wages could be offset by increase labour supply. The relative shifts in labour demand for the more or less highly skilled could potentially affect either relative wages or employment levels. In a country where wages are comparatively rigid, the impact of technological change may result in declining relative employment for less skilled workers rather than changing relative wages. This is sometimes called the “ skill biased technological change” hypothesis since technologies are thought to favour the outcomes of skilled workers relative to unskilled workers. New employment arrangements At the same time, many firms changed the way they manage their workforces. “ Downsizing” and “ contracting out” became dominant buzzwords of the 1990s. Self-employment, which by nature is a less stable employment relationship than traditional “ paid” work, grew substantially as a fraction of total employment in Canada. Change in the way firms manage their workforces may have affected labour market performance. Betcherman and Lowe (1997) argue that there are six threads that run through employers’ search for a new model of management: global integration, technological change, innovations in work organization, business rationalization, the pursuit of high performance workplaces, and new forms of labour relations. To date, we know little about the extent to which most of these phenomena are taking place in Canadian firms, let alone their impact on labour market outcomes or the income of Canadians. Osberg, Wein and Grude (1994) provide interesting case study analyses describing how new workplace management principals have affected employment relationships at some Nova Scotia employers. The rapidly increasing human capital of Canadian workers While the above mentioned factors reflect changes in labour demand, labour supply changes were perhaps equally as dramatic. Most notably, the supply of highly educated workers continued to increase at a remarkable rate in the 1990s, spurred on by the expansion of the post-secondary
Analytical Studies Branch – Research Paper Series
-2-
Statistics Canada No. 11F0019 No. 184
education system that started in the 1960s, and the retirement of much less educated cohorts of workers (Riddell, 1995). Within the context of a general rise in the educational attainment of workers, there has also been a shift in relative educational levels. In particular, the educational attainment of older workers and younger women has been rising faster than that of young men. Given the strong association between education and earnings, shifts in relative educational attainment will contribute to shifts in relative wages. The experience levels of Canadian workers also rose during the 1990s, as the workforce aged due to the maturation of the baby boom generation. Because of rising education levels and experience, the human capital embedded in Canadian workers rose significantly. Finally, historical increases in women’s participation rates and their rapidly rising educational attainment reflect important social changes in Canada that potentially have an impact upon the supply of skilled workers, the relative earnings of men and women, and the income of families. Beyond the direct impact of a more educated, experienced workforce on relative and real earnings, Acemoglu (1998) and Beaudry and Green (1999) argue that firms invested more in technology in order to take advantage of the rising educational attainment of workers. This hypothesis argues that the increased educational attainment of the workforce has encouraged firms to alter their production methods to make use of these higher skills. If firms invest strongly in technology, this tends to push down demand for unskilled workers. This argument stands in contrast to the skill biased technological change argument where it is technology that is the prime mover. Changes in the institutional landscape The institutional landscape was also altered during the 1990s. Substantial revisions to Employment Insurance (EI) could potentially influence structural unemployment, employment and the income of low-income Canadians relative to earlier decades. The major revisions to the social assistance system in many provinces in the late-1980s (expansion) and mid-1990s (contraction) and the introduction of the child tax benefit system by the federal government could also affect incomes and employment among low-income Canadians. Minimum wages have changed little since the early 1980s, but the unionization rate has fallen (Sargent, 2000). Fortin and Lemieux (1997) found that declining real minimum wages, falling unionization, and deregulation were important determinants of rising wage inequality in the U.S. Aggregate demand during the 1990s Finally, all these changes took place within a national context of weak aggregate demand which characterized the economy for most of the decade. Real gross domestic product grew at almost 3% per year over the 1980s, but at only 1.8% per year from 1990-1998. Associated with weak demand was weak employment growth rooted in stagnant hiring. The final years of the decade were years of strong growth however, with the economy posting 4.5% and 4.7% real growth rates in 1999 and 2000 respectively. Canada’s performance in 2000 was impressive compared with other G7 nations, with only the U.S. posting faster expansion (Cross, 2001). Growing aggregate demand translated into growing employment, as the unemployment rate fell to 6.8%, its lowest annual level since the current labour force survey began in 1976. Mishel, Bernstein and Schmitt (2001a) argue that persistently low unemployment was a major factor behind faster wage and
Analytical Studies Branch – Research Paper Series
-3-
Statistics Canada No. 11F0019 No. 184
income growth in the U.S. of the late 1990s. Persistently low unemployment, it is argued, helped raise growth in productivity and gave workers opportunity to shift away from contingent jobs. Changes in aggregate demand over a business cycle influence both relative and real wages. Earnings inequality and low-income rates tend to increase in economic contractions and fall in expansions. This is an important consideration when studying distributional trends. Often one is attempting to determine if there are structural (i.e. longer-term and more permanent) changes in inequality and relative earnings. It is necessary to abstract from cyclical changes in aggregate demand and its effect on inequality to achieve this.
Changes in income and earnings inequality Few labour market issues have evoked as much interest in the past two decades as earnings and income inequality. Differing trends among countries, particularly between Canada and the U.S., often lead to confusion regarding Canadian outcomes. Adding to the confusion is the fact that trends may differ depending upon whether one is referring to individual earnings inequality, inequality in family market earnings, or inequality in family disposable income. This review focuses on all three levels, and contrasts outcomes with those of the U.S. Why focus on all three levels? Earnings inequality (wages from jobs) provides a window on how the labour market is distributing employment earnings among individual workers. It is primarily a reflection of the interaction of labour supply and demand forces, as well as the effect of institutional features of a labour market (e.g. labour legislation, minimum wage levels, social assistance and Employment Insurance (EI) regulations, etc.). Family market earnings inequality simply places individual workers into families, determines family earnings, and adds investment earnings. This relates more to the "welfare" of Canadians than individual earnings inequality (which is more of a reflection of events in the labour market), since it is concerned with family outcomes. Finally, inequality in disposable family income is the level at which most analyses concerned with welfare or well-being focus. Social transfer income (e.g. social assistance, public pensions, EI benefits, tax credits, etc.) is added to market income, and the distributive effects of taxes are included. This provides a measure of how all money income available to Canadian families is distributed. Changes in the inequality at this level reflect the distributional outcomes driven by labour market forces, government transfer and taxation policies, and social forces affecting family formation. Earnings inequality rose in a number of countries, including Canada, U.S. and the U.K. between the 1970s and 1980s. There was a substantial fall in the earnings of lower-paid workers, particularly males, while the earnings of the higher-paid either remained stable or increased (Beach and Slotsve, 1996; Morissette, Myles and Picot, 1994). The gini coefficient is a wellknown summary measure of inequality. This measure takes on values between 0 and 1, a higher value indicating that earnings inequality is greater, and the variation in earnings among workers has increased.
Analytical Studies Branch – Research Paper Series
-4-
Statistics Canada No. 11F0019 No. 184
Table 1 Inequality in Individual Earnings, Family Market Earnings and Disposable Family Income 1974 Individual Employment Earnings* Canada Men Women Both U.S. Men Women Both
Gini Coefficient 1985
1995
1997
0.351 0.407 0.406
0.392 0.426 0.428
0.402 0.421 0.423
0.397 0.411 0.418
0.377 0.425 0.437
0.413 0.428 0.446
0.456 0.441 0.467
0.456 0.439 0.466
Family Market Earnings** Canada U.S.
0.389 0.421
0.402 0.459
0.427 0.501
0.423 0.502
Family Disposable Income*** Canada U.S.
0.303 0.327
0.293 0.355
0.291 0.379
0.292 0.392
*
Includes paid employment plus self-employment earnings. Includes persons aged 18-64 who earned more than $500 during the year. ** Includes employment earnings plus investment income. Family earnings are adult equivalent adjusted (to account for economies of scale associated with larger families) and the individual is the unit of analysis. Based on economic families. *** Based on economic families. Family income is adult equivalent adjusted, and each individual is assigned the adult equivalent adjusted family income. The individual is the unit of analysis. Includes employment and investment earnings, pension, social transfer income and taxes paid. Source: Wolfson and Murphy (2000).
Earnings inequality tends to rise in recessions and decline in economic expansions, and hence when seeking longer-term trends, the years considered should be in the same part of the business cycle (or the cyclical variation removed from the series). In Canada, the gini for individual employment earnings among men rose from 0.351 in 1974 to 0.392 in 1985 (up 11.7%), then a lesser increase to 0.402 in 1995 (up 2.6%) (Table 1). Among women the changes were less significant, up 5.4% between 1974 and 1985 and steady between 1985 and 1995. With the improving economy between 1995 and 1997, ginis fell for men and women. In contrast, earnings inequality for men in the U.S. rose steadily, up 9.5% between 1974 and 1985 and up 10.4% between 1985 and 1995. As in Canada, changes in earnings inequality among American women was more muted. However, with the rise in the share of women working, the increased educational attainment of young women, and the continued movement of younger women into traditionally “ male” occupations, women and men are becoming increasingly integrated in the labour market. From a labour market perspective, the distinctions between them, while still very evident, are declining. Hence, it is useful to view earnings inequality from the perspective of all workers, men and women combined, and think about a single labour market. Earnings inequality among all workers rose through the 1980s, but increased little in the 1990s in Canada. The gini for all workers rose from
Analytical Studies Branch – Research Paper Series
-5-
Statistics Canada No. 11F0019 No. 184
0.406 to 0.428 between 1974 and 1985. By 1995, however, a year that is roughly in the same position in the business cycle as 1985, no increase was evident, and the gini had fallen slightly by 1997. Although individual earnings inequality rose in the 1980s, there is little evidence of strong increases in Canada in the 1990s (Green, 1999; Picot, 1998; Wolfson and Murphy, 2000). In the U.S., on the other hand, earnings inequality continued to increase to 1997 (Table 1). However, earnings inequality represents a very broad and general indicator of distributional features of the labour market. No increase in inequality, as in Canada’s case during the 1990s, does not imply that there were not significant redistributional events affecting particular groups, and substantial shifts in relative wages. Such movement can be offsetting, so that while there are important changes in the earnings of one group relative to another (say younger men compared to older men, or women compared to men), overall inequality may not be seen to increase. These and other trends offset each other so that among all workers, the distribution of employment earnings changed little in the 1990s (Picot, 1998). When earnings are aggregated to the family level, inequality is seen to continue increasing in the 1990s. Data in Table 1 are adult equivalent adjusted, so change in the size of families is accounted for and the results include individuals and families. The gini for family market earnings inequality rose from 0.389 in 1974 to 0.402 in 1985, and continued to rise to 0.427 in 1995 (and 0.423 in 1997). From a welfare perspective, this is perhaps a more useful measure of earnings inequality than is individual earnings. The differences in outcomes between family and individual earnings inequality primarily reflects changes in marriage and family formation patterns. Increasing numbers of single parent families (who tend to have low family earnings) and an increasing tendency for "like" to marry "like" (i.e. low earners marrying low earners, high marrying high) resulted in rising family market earnings inequality, even though individual earnings inequality did not rise in the 1990s (Zyblock, 1997; Wolfson and Murphy, 2000). When considering income measures only, as we are here (i.e. excluding references to wealth and other aspects of economic well-being) perhaps the best reflection of changes in the relative “ economic welfare” of Canadians is inequality in family disposable income. This reflects the distribution of total money income available to families, including earnings, pensions, social transfers and other market and non-market sources of income, and also accounts for the effect of personal income taxes on inequality. There has been little change in family income inequality throughout the 1970s, 1980s or most of the 1990s in Canada. The gini was 0.303 in 1974, 0.293 in 1985, and 0.292 in 1997 (Table 1). Generally speaking family income inequality has changed little during the past 30 years. This is because employment earnings losses among lower income families were largely replaced by increases in social transfer payments (e.g. social assistance, EI benefits, rising public pension income received, child tax credit, etc.) throughout the 1980s in particular, preventing inequality from rising. More recent data does suggest that during 1997 and 1998 there may have been an increase in family income inequality. Chart 1 shows gini coefficients from 1989 through 1998. For families only (excluding unattached individuals), the level of disposable income inequality was steady from 1989 through 1995 at a 7-year average of 0.294, and rose to between 0.312 and 0.315 in 1997-1998. Similarly, disposable income inequality among unattached individuals was steady between 1989 and 1995 at an average of 0.337, rising to an average of 0.354 between 1996 and 1998. However, these figures are not adult equivalent adjusted, and changes in
Analytical Studies Branch – Research Paper Series
-6-
Statistics Canada No. 11F0019 No. 184
data collection over this period may also have affected these results1. More research is needed to better understand the change in family income inequality during the late 1990s. Chart 1: Inequality of Family Disposable Income, GINI coefficients 0.4
Unattached Individuals
gini
0.35
0.3 Economic Families with Two or More Members
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
0.25
Year
Statistics Canada Cat. No. 75-202-RPE. Income in Canada. 1998.
In the U.S. on the other hand, family income inequality has increased steadily. The gini rose from 0.421 in 1974, to 0.459 in 1985, and 0.502 by 1997 (Table 1). Faced with a similar increase in earnings inequality as in Canada during the 1980s, the U.S. transfer system did not respond in the same manner to prevent increases in inequality (Blank and Hanratty, 1993). Much of the common perception in Canada that family income inequality has increased during the past three decades may well stem from these American results. As noted earlier, inequality and relative wages can change significantly over the business cycle. In particular, during periods of low unemployment and relatively high labour demand late in an economic expansion, the relative position of low wage workers can improve. Mishel, Bernstein and Schmitt (2001a) examine wage inequality in the U.S. in the 1990s. They found that low earners were closing the gap with middle and working class workers since 1995, while the top continued to pull away from the middle. Regarding family income inequality, the top continued 1
Adult-equivalent adjustment is a method of adjusting family income so that family size, and the economies of scale associated with larger families, are accounted for. The data in Table 1 are adult-equivalent adjusted, those in Chart 1 are not. If family size is changing over time, it is desirable to “ adult-equivalent adjust” the results. Regarding data sources, the data in Table 1 come from the Survey of Consumer Finances only. It was discontinued after data year 1998. In Chart 1 data up to 1995 are from the Survey of Consumer Finances, while from 1996 to 1998 data are from the Survey of Labour and Income Dynamics.
Analytical Studies Branch – Research Paper Series
-7-
Statistics Canada No. 11F0019 No. 184
to pull away from the middle and the middle pulled away from the bottom in the second half of the 1990s. They argue that persistently low unemployment in the later 1990s, coupled with rising minimum wages in that decade (after declines in the 1980s) contributed to improvement at the bottom of the distribution, while globalization, deunionization and the shift to low-paying industries have kept wages at the top growing. In contrast, distributional changes appear unrelated to technological change. One piece of evidence they find is that high tech sector employers have not been the source of growth in employment or been wage leaders (in terms of growth) in the U.S.
Changes in relative earnings Changes in inequality are generally related to changes in relative earnings among groups, for example, of older workers compared to younger, and more versus less skilled. However, the two concepts are not synonymous. Inequality trends include not only changes in relative wages among groups (defined by age, education, gender, etc.), but also within groups. And changes in “ within” group inequality have been a very important, and often ignored, part of the inequality story. Furthermore, aggregate inequality measures, such as the ginis presented in the earlier section, can show little change, while beneath this apparent stability there can be important, and offsetting, movement in relative wages of various groups. Here, we focus on the changes in the relative earnings of older and younger workers, the more and less educated, and between men and women. These are topics that have received considerable attention during recent decades. Declining real earnings among young workers, particularly men Cross-sectional data clearly indicate an increasing earnings gap between younger and older workers, particularly among men, during the 1980s and 1990s. This was related mainly to a fall in the real earnings of younger male workers (Betcherman and Morissette, 1994; Picot, 1998). Real weekly earnings of younger men (under 35) working full-time fell through both the recessions of the early 1980s and 1990s. Modest increases for 18-24 year olds during the 1980s expansion, and for both 18-24 and 25-34 year olds during the late 1990s expansion failed to fully offset these declines. Meanwhile, earnings rose for older men (aged 45-64) further increasing the wage gap (Chart 2). For women, earnings declined only for the youngest group of 18-24 year olds (Chart 3).2 Rising relative wages for older men might suggest an increase in returns to experience. Young workers may start out with relatively low earnings, but eventually “ catch up” as they age and gain experience. However, this does not appear to be the case. Beaudry and Green, 2000, found that if one examines successive cohorts of men as they entered the labour market and gained experience, the age-earnings profile has been shifting downward for more recent cohorts. Wages have fallen for young men as they enter the labour market, and their wages do not catch up to those of earlier cohorts as they gain experience. This decline in wages is observed for both the more and less highly educated. More recent work using data from a longitudinal graduates survey
2
A consistent series for wages up to 2000 is not yet available. For the indices in Charts 2, 3 and 6 we use the Survey of Consumer Finances (SCF) data up to 1997, then estimate expected growth in the wage series from 1997 to 2000 using Labour Force Survey (LFS) data.
Analytical Studies Branch – Research Paper Series
-8-
Statistics Canada No. 11F0019 No. 184
also uncovers similar trends for male bachelors graduates, although the magnitude of the decline has been slight over the 1990s (Finnie, 2001). Little is known regarding the cause of the declining real and relative (to earlier cohorts) earnings of young workers, males in particular. Changes in supply do not explain the drop since the supply of younger workers has been falling, not rising, both in absolute terms and relative to the overall supply of labour. This would tend to increase, not reduce relative wages. Proponents of the skill biased technological change hypothesis draw a connection between increasing returns to skill, and changes in the distribution of earnings by age. They argued that skill biased technological change increased the demand for skills of all types and that rising male earnings inequality was consistent with increasing returns to experience or age (Juhn, Murphy and Pierce, 1993; Katz and Murphy, 1992; Bound and Johnston, 1992; Davis, 1992). As was the case of the role of technological change in explaining the shifts in relative earnings by education level, some argue the role of technological change in explaining falling real wages among young men may also have been overstated. Reviews by Green (1999) and Sargent (2000) argue that the changes in labour market outcomes and industrial structure that prevailed in the 1990s are not necessarily those that one might expect to see in an economy where technological change was driving structural change. Changes in information technology have clearly had an impact on labour market outcomes for workers, and such change is part of the evolution of the 1990s labour market. But there are important phenomenon that cannot be explained and are at times inconsistent with the belief that technology is the sole, or perhaps even primary, driving force. For example, earnings inequality in Canada has risen and continues to rise among men, but has changed little among women (Picot, 1998). Green (1999) asks how one can reconcile this with technological change, which is presumably continuing into the 1990s, and should affect women as well as men. Also, as discussed above, wages have declined across cohorts—increasing returns to experience have not taken place within cohorts—so the argument that increasing wage premiums for experience is consistent with the effects of technological change holds little validity, (Beaudry and Green, 2000b). And why have earnings of young men been falling if technological change in the shift to the knowledge based economy were boosting demand for skilled labour? It is among the young that computer-related skills are the highest. In the U.S. where the increase in the wage premium (for education and experience) has been substantial, the increase accounts for only about 45% of the overall rise in inequality, the remainder is found within groups of similar educational attainment and experience. (Mishel and Bernstein, 2001b) Hence, even if the rise in the education and experience wage premium were fully driven by technology, that leaves much of the overall increase in earnings inequality unexplained. However, technology is likely only one of several factors which explain this premium, others being a drop in the real federal minimum wage over the 1980s and a continued decline in unionization in the U.S. (Mishel and Bernstein, 2001b). These and other inconsistencies in the evidence supporting the skill biased technological change hypothesis have encouraged researchers to look to other explanations, such as the impact of changes in institutions as argued by Fortin and Lemieux (1997).
Analytical Studies Branch – Research Paper Series
-9-
Statistics Canada No. 11F0019 No. 184
It may be that changes in labour supply partly explain why young males in particular saw their relative earnings fall. This relates in particular to the rising education of older workers and younger women. In 1981, 16.4% of men 25 to 34 years old had a degree, compared with 11.8% of men 45 to 54. By 1995 this gap had disappeared, each group rising to 22.2% and 22.5% respectively. Both age groups had become much more highly educated, but more so for older men. The gap did not close to the same extent for women, largely because of the dramatic increase in the educational attainment of the young: 16.6% of younger women held a degree in 1981 compared to 6.9% of older women. By 1995, this had risen to 25.3% among the young (higher than for men), and 19.2% of older women. This rapid increase among workers 45 to 54 was related to the rapid increase in the university enrolment rates during the 1960s and early 1970s, and the fact that these cohorts are now moving into the 45-54 year old age group, replacing much less educated cohorts. Kapsalis, Morissette and Picot (1999) conclude that the changing relative educational levels of the young and old accounted for about one-quarter of the rising cross-sectional earnings gap between younger and older workers during the 1980s, and even a larger share during the 1990s. We discuss the declining educational advantage of young men in more detail later in this paper. Chart 2: Indexed Median Earnings, Full Time Male Workers, $1992
1.1
1.05 45-64 1
35-44
0.95 25-34 0.9
0.85
0.8 18-24 0.75
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
0.7
Source: Survey of Consumer Finances 1980-1997, Labour Force Survey 1997-2000.
Analytical Studies Branch – Research Paper Series
- 10 -
Statistics Canada No. 11F0019 No. 184
Chart 3: Indexed Median Earnings, Full Time Female Workers, $1992
1.3
1.25 45-64 1.2
1.15 35-44 1.1
1.05 25-34 1
0.95
0.9 18-24 0.85
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
0.8
Source: Survey of Consumer Finances 1980-1997, Labour Force Survey 1997-2000.
No marked increase in the wage premium for higher education in Canada among all workers, a rising premium for younger workers When considering all workers, the basic facts suggest that through the 1980s and mid- 1990s there was no overall increase in the returns to education in Canada (Morissette, Myles and Picot, 1994; Bar-Or, Burbidge, Magee and Robb, 1993; Beach and Slotsve, 1996). This is in contrast to developments in the U.S., where the wage premium among the more highly educated has risen (Juhn, Murphy and Pierce, 1993; Freeman and Needels, 1991). A developing literature examines the role played by technological change in explaining the phenomenon of increasing relative wages for the more educated in the U.S. The skill biased technological change hypothesis states that technological change has increased the demand for more highly skilled workers. This should result in increasing relative wages for this group (relative to the less skilled), unless wages were highly rigid, causing a shift in relative employment, or supply of educated workers rose enough to offset demand. Interestingly, neither relative wages nor employment rates appear to have changed significantly among the more highly educated in Canada among workers as a whole. The relative employment to population ratios of the university to high school educated suggests that there has been no relative gain among the more highly educated in terms of the likelihood of being employed, although the likelihood of being employed has fallen for those with less than a high school education (Sargent, 2000). It is not so much that the relative position of the very highly educated improved, rather the position of the least educated deteriorated. An explanation for this stable pattern of relative earnings Analytical Studies Branch – Research Paper Series
- 11 -
Statistics Canada No. 11F0019 No. 184
in Canada that is consistent with the skill biased technological change hypothesis, is that the relative supply of highly educated workers increased in Canada at a rate that offset the increase in the relative demand. Increases in supply offset increased demand caused by technological change, resulting in little change in the relative labour market status of skilled and unskilled workers (Freeman and Needels, 1991; Murphy, Riddell and Romer, 1998). This is a commonly accepted explanation of the difference between Canada and the U.S. with respect to changing relative wages of the highly educated. A second hypothesis recently put forth by Beaudry and Green (1999) argues that the role of technological change in causing the relative earnings gap in the U.S. may be overstated. They argue that firms invested more in technology in order to take advantage of the rising educational attainment of workers. This model, sometimes called the “ endogenous technical choice” hypothesis argues that the increased educational attainment of the workplace has encouraged firms to alter their production methods to make use of these higher skills. When firms invest strongly in technology, this tends to push down demand for unskilled workers. In this scenario, it is the rising educational attainment of the workforce that is the principal cause of many of the changes in the relative wages. This possibility is also stressed in theoretical work by Acemoglu (1998). A third explanation that has received less attention in Canada is that institutional factors underlie the changes in real wages. Fortin and Lemieux (1997) found that declining real minimum wages, falling unionization, and deregulation were important determinants of rising wage inequality in the U.S. The relative merit of these proposed explanations continues to be debated. Most recently Beaudry and Green (2000) compare the changing structure of wages in the U.S. and Germany. Examining the past 20 years, they point out that while the educational composition of the workforces in the two countries evolved similarly, the wage premium for education rose more in the U.S. One explanation for this that was put forth by the OECD (1994) was that Germany had relatively rigid wages, and as a result “ skill biased technological change” manifested itself in rising unemployment rates among the less skilled. Beaudry and Green offer the alternative hypothesis that the difference is because the U.S. under accumulated physical capital relative to human capital, while Germany invested in a more balanced manner. The faster accumulation of physical capital in Germany helped ensure sufficient demand and productivity gains for lower skilled workers that kept their wages from declining. The debate continues. In any case, stating that there has been no increase in the education premium for workers in Canada obscures relative changes within some groups. Bar-Or, Burbidge, Magee and Robb (1993) examined wages by education group from 1971-1991 using Survey of Consumer Finances data and found that returns to higher education did not rise in the 1980s except for among the youngest age group (those with experience levels of 1 to 5 years), who saw modest increases. Furthermore, recent data from the Labour Force Survey shows that young university educated workers have seen real and relative (relative to the less educated) earnings increases in recent years. Chart 4 shows average real weekly wages for the 20-34 year old age group from 1997 (the first year the LFS collected wage information) to 2000. While other groups stagnated in earnings, men in particular with university degrees enjoyed small wage increases over these years.
Analytical Studies Branch – Research Paper Series
- 12 -
Statistics Canada No. 11F0019 No. 184
Chart 4: Average Real Weekly Wages, 1997-2000 , Age 20 to 34 800
700
600
$1992
500 1997 1998 1999 2000
400
300
200
100
0 men, less than high school
men, high school
men, diploma or certificate
men, university degree or higher
women, less women, high than high school school
women, diploma or certificate
women, university degree or higher
Source: Labour Force Survey.
The shifting relative labour market position of men and women The changing relative position of women and men is one of the more interesting labour market phenomenon of the past two decades, and one that may not be receiving the attention it is due. The changing relative position of women has several dimensions. Young women have rapidly increased their educational levels relative to men. They are occupying more jobs, and while wages remain lower for women, the gap is closing. Generally speaking, the likelihood of being employed has increased for women and declined for men, while that of being unemployed has risen for men, and fallen for women. The offsetting trends between these two groups are seen in earnings levels, earnings inequality, the likelihood of being employed (employment population ratios) and the likelihood of being unemployed. We discuss each of these in turn below. Educational attainment is rising faster for women There has been an increasing educational attainment of women, both in real terms, and perhaps more significantly, relative to men. For example, between 1976 and 1998 the fraction of women in the labour force with a university degree was increasing at 7.7% per year, compared to 4.2% for men. The result is that the gap between men and women regarding the share holding degrees disappeared by 1998. In fact, the educational advantage held by young men (over younger women and older men) in 1976 had more than disappeared by 1998 at which time a larger share of young (25-34) women than men held degrees (Chart 5).
Analytical Studies Branch – Research Paper Series
- 13 -
Statistics Canada No. 11F0019 No. 184
Chart 5: Proportion of Population With University Degrees 0.3
0.25
0.2
men age 25-34 women age 25-34
0.15
men age 45-54
0.1
0.05
0 1976
1990
1998
Year
Source: Labour Force Survey
Since human capital is a major determinant of labour market outcomes, it should not be too surprising that the relative earnings, employment and unemployment of women are improving. However, there are likely other factors at play. Picot and Heisz (2000) noted that even after controlling for change in human capital characteristics (education and age), employment and unemployment outcomes deteriorated for men, but not for highly educated women, and relatively little for other women. And as will be shown below, even after controlling for broad education and age (experience) levels, women’s earnings are rising faster than men's during the first half of the 1990s. The male-female earnings gap is declining While aggregate data indicate that women continue to earn less than men, the gap has been narrowing. The ratio of annual earnings of women to men working full-year full-time rose from 58.4% in 1967 to 72.5% in 1997 (Survey of Consumer Finances data). Annual earnings are not the best indicator however of relative wages, since they do not fully control for weeks worked in the year or hours worked per week which are known to be higher for men. Drolet (2000) finds that the raw hourly wage ratio of women to men was 80% in 1997, but that increased to 89% after controlling for a host of human capital, productivity related and industrial and occupational characteristics.
Analytical Studies Branch – Research Paper Series
- 14 -
Statistics Canada No. 11F0019 No. 184
And over the 1990s, the aggregate earnings of women have been increasing much more quickly than that of men. The weekly earnings of female full-time workers displayed a significant increase in the 1990s compared to the 1980s, rising 13% between 1989 and 2000 (Chart 6). Weekly earnings were stable among comparable male workers. Controlling for age, education, full-time, part-time status and industry, produces much the same results. Kapsalis, Morissette and Picot (1999) ran wage equations for men and women (separately) in different age groups and computed expected weekly earnings controlling for industry, region, and full-time/part-time status. During the 1989-95 period, the growth in expected weekly earnings was greater for women than men (except for the university educated) (Chart 7). Basically, men’s weekly earnings were declining and women’s were rising. Evidence based on annual earnings suggests this phenomenon is independent of where in the earnings distribution one is located. The growth in annual earnings of women has far outstripped that of men across the entire earnings distribution. Between 1986 and 1995, two years in roughly comparable positions in the business cycle, women’s annual earnings rose 10.6%, and men’s fell by 1%. Except for women at the very bottom of the earnings distribution (where earnings rose by only 0.4%), earnings growth was between 9% and 11%. Men’s earnings fell across the entire earnings distribution, except among those in the top two deciles, where there were earnings gains (Chart 8). Chart 6: Indexed Median Earnings, Full-Time Workers
1.2
1.15
Women
$1992
1.1
1.05
All
1 Men
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
0.9
1980
0.95
Source: SCF 1980-1997, LFS 1997-2000.
Analytical Studies Branch – Research Paper Series
- 15 -
Statistics Canada No. 11F0019 No. 184
There are numerous possible reasons for these phenomena. Women are moving into higher paid occupations and fields of study at an increasing rate, and such effects would not necessarily be controlled for in the broad type of controls used in the work reported here. The wage gap has been narrowing for male and female graduates, and the higher the education level, the lower the gap. For the latest cohort of graduates in their study (1990), Finnie and Wannell (1999) found that the male/female wage gap for doctoral graduates had disappeared, at least during the five years following graduation. Chart 7: Growth in Average Weekly Earnings, 1989-1995 20 women 25-34 15
% growth 1986-1995
10
women 45-54
5
0
-5
men 45-54
-10
-15
men 25-34
-20 elem
high sch
some ps
dipl/cert
univ
Source: Kapsalis, Morissette and Picot, 1999.
Chart 8: Growth in Average Annual Real Earnings, 1986-1995 15
% growth 1986-1995
10
5 men 0 women -5
-10
-15 all
1
2
3
4
5
6
7
8
9
10
deciles
Source: Kapsalis, Morissette and Picot, 1999. Analytical Studies Branch – Research Paper Series
- 16 -
Statistics Canada No. 11F0019 No. 184
Employment indicators improving for prime aged women relative to men Earnings growth was largely positive for women and stable or declining for men during the 1990s. This dichotomy is reflected in employment and unemployment outcomes as well. There may be special conditions influencing the outcomes of younger workers (changing employment patterns among students) and older workers (increasing early retirements) that could affect the relative change between men and women. Hence, we focus on prime aged workers, aged 25-54. Labour force statistics suggest that the position of prime aged men has deteriorated between the late 1980s and late 1990s, while that of women has improved. We computed average employment and unemployment rates for the 1987-1989 period, and compared them to average rates for the 1998-2000 period. Both periods are roughly at a comparable position in the business cycle (approaching the cyclical peak) and hence the comparisons represent a structural change between the decades, not just cyclical variation. Labour force participation fell for men (by 2.7%), and rose for women (7.2%) (Table 2). The unemployment rate for men remained largely unchanged (falling 0.1 percentage points) but fell substantially for women (1.5 percentage points). Hence, while unemployment was higher among women than men during the 1980s, by 1998-2000 this was reversed. The employment population ratio fell for men (2.6%) and rose for women (8.9%). The one statistic which worsened for women was unemployment duration, which rose by 11.7% among women, but declined marginally among men (2.8%), comparing 1984-1986 to 1994-1996 which are the most recent years available. These changes to duration were more than offset by increases in the incidence of unemployment among men, and declines among women (by 15.7%). There are certainly many reasons for this change in relative labour market position, including changes in the types of jobs held by women, and the industries in which they work. However, little is known regarding these relative shifts in earnings, employment and unemployment, and this remains one area requiring further research.
Analytical Studies Branch – Research Paper Series
- 17 -
Statistics Canada No. 11F0019 No. 184
Table 2: Labour Force Statistics, 25 to 54 Year Olds Only 1987-1989
1998-2000
% change difference
Labour Force Participation Rate
All Men Women
83.2 93.6 72.9
84.6 91.1 78.1
1.6 -2.7 7.2
1.3 -2.5 5.2
Unemployment Rate
All Men Women
7.2 6.6 7.9
6.4 6.5 6.3
-10.2 -2.0 -19.5
-0.7 -0.1 -1.5
Employment to Population
All Men Women
77.3 87.4 67.2
79.1 85.1 73.1
2.4 -2.6 8.9
1.9 -2.2 6.0
1984-1986
1994-1996
a
% change difference
Inflows to Unemployment *
All Men Women
1.95 1.73 2.25
1.86 1.84 1.89
-4.4 6.1 -15.7
-0.1 0.1 -0.4
Duration of Unemployment (Months)*
All
4.3
4.4
2.5
0.1
Men Women
4.6 3.8
4.5 4.3
-2.8 11.7
-0.1 0.4
a: fraction of the labour force entering unemployment *1996 is the most recent year available for inflows to and duration of unemployment. Source: Labour Force Survey
Low-income trends in the 1990s While inequality measures indicate how income is distributed among all Canadians, low-income measures focus on the families at the bottom of the income distribution. Traditionally, analysis focuses on the share of the population below some low-income cutoff (the low-income rate), and asks whether this rate is rising or falling, and why. But it is also important to know whether the families in low-income are better or worse off than they were in the past. This is determined by the low-income gap (the income gap between the low-income cutoff and average family income among low-income families). Finally, another important dimension of low-income is its persistence. How long do families tend to stay in low-income once there. If families are “ trapped” in low-income for many years, this represents a more serious issue than if families move in and out of low-income relatively quickly. This section discusses trends in these three components—the low-income rate, the low-income gap, and the persistence of low-income among Canadian families.
Analytical Studies Branch – Research Paper Series
- 18 -
Statistics Canada No. 11F0019 No. 184
The low-income rate There are a number of low-income cutoffs in use. If the income of a family fall below the cutoff, they are said to be in the low-income population. Here we use Statistics Canada’s LICO (lowincome cut-off) based on after-tax and transfer income (i.e. disposable income). In 1998 dollars, the post-tax LICO in a large urban centre was $14,510 for a single person, and $27,890 for a family of four. How has the low-income rate fared in the 1990s relative to the 1980s? In answering this question it is important to separate cyclical variation in low-income from longer-term trends. Lowincome rates always rise in recessions, and fall during economic expansions. Chart 9 contains the rate of low-income among all Canadians based on after-tax income for the period 1980 to 1998. Also plotted in Chart 9 is the unemployment rate. Cyclical movements in the low-income rate mirrored cyclical changes in the unemployment rate, at least up to 1994 (the correlation coefficient over 1980 to 1994 was 0.85). Following that year, trends in low-income diverged from the unemployment rate, such that 1996 and 1997 were peak years for low-income despite improving labour markets. The low-income rate recovered somewhat between 1997 and 1998, but by 1998 had not fully returned to the level expected based on the business cycle. Chart 9: Low Income Rate (left axis) Unemployment Rate (right axis) 15
14 low-income after-tax rate
13
14 12 11 10 12 9 11 8
unemployment rate
unemployment rate (%)
low-income rate (%)
13
7
10
6 9 5
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
4 1980
8
Year
Source: Labour Force Survey, Statistics Canada Cat. No. 13-592. Low-income After Tax, 1997.
Table 3 provides data on persons in low-income (based on the post-tax LICO) for 1981, 1988, 1989—the cyclical peak year during the 1980s, for 1993—when the overall economic recovery began to take hold; and for 1998. As shown, the overall low-income rate for all persons in 1998 was 12.2%, down from 13.1% in 1993, but 1.1 percentage points above the 11.1% level of 1988, a year that is roughly in a comparable position of the business cycle. Overall the low-income rate Analytical Studies Branch – Research Paper Series
- 19 -
Statistics Canada No. 11F0019 No. 184
was somewhat higher during the 1990s than 1980s. It rose for four years following the end of the recession (1992), peaking in 1996. This increase occurred in spite of a slowly recovering economy, and falling unemployment. During the 1980s the increase lasted only one year following the recession end (1982), peaking in 1983. Different groups were affected in different ways. Low-income rates for seniors—already very low for seniors in families—fell over the 1990s. In contrast, low-income rates among the 18-64 population were higher in 1998 than in 1988, a year in roughly the same position in the cycle. This rise was observed both among singles (increasing 2.9 percentage points between 1988-98) and families (increasing 1.3 percentage points between 1988-98). The low-income rate of children dropped from the cyclical high of 17.2% in 1996 to 13.8% in 1998. However, the rate in 1998 was still above the 12.5% posted in 1988, a comparable year in the business cycle. The low-income rate was higher in the late 1990s than late 1980s. The opposite trend was observed during the 1980s; the children’s low-income rate fell from 12.5% in 1981 to 11.8% in 1989 (comparing cyclical peaks). Table 3 Low-Income Rate: Persons in Low-Income After Tax (% of Group) and % Composition of Low-Income Population 1981
1988
1989
1993
1998
% of Low-income Population (1998)
All
11.7
11.1
10.2
13.1
12.2
100
Men Women
10 13.5
9.5 12.6
8.8 11.5
12 14.3
11.4 13
46.1 53.9
Elderly In Families Single
7.2 49.1
4.1 30.6
3.1 27.3
3.3 26.3
3 20.8
8.7 2 6.7
Children (
